US7734107B2ExpiredUtilityPatentIndex 63
System and method for performing wavelet-based texture feature extraction and classification
Est. expiryFeb 24, 2026(expired)· nominal 20-yr term from priority
Inventors:LI PINGSHAN
G06F 16/5862G06V 10/431G06V 10/435G06V 10/52
63
PatentIndex Score
4
Cited by
13
References
11
Claims
Abstract
A system and method are disclosed for performing wavelet-based local texture feature extraction and classification procedures. Image data is initially provided to include a query image and a series of test images. A feature detector calculates image parameters corresponding to the image data. The image parameters include mean absolute values, variance values, and texture angles. The feature detector utilizes the image parameters to calculate distance values that represent texture similarity characteristics between the query image and each of the test images. The feature detector then evaluates the distance values to determine one or more matching images from among the test images.
Claims
exact text as granted — not AI-modified1. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said image parameters including coefficient mean absolute values for each of said subbands, a coefficient mean absolute value μ for a subband LH being calculated utilizing a formula:
μ
LH
(
i
)
=
1
MN
∑
m
=
1
M
∑
n
=
1
N
W
LH
(
i
)
[
m
,
n
]
where LH(i) is said subband LH at a level i, W is a wavelet coefficient, m is a coefficient row, n is a coefficient column, M is equal to a total number of coefficient rows, and N is equal to a total number of coefficient columns.
2. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said image parameters including coefficient variance values for each of said subbands, a coefficient variance value σ for a subband LH being calculated by utilizing a formula:
σ
LH
(
i
)
2
=
1
MN
∑
m
=
1
M
∑
n
=
1
N
(
W
LH
(
i
)
[
m
,
n
]
-
μ
LH
(
i
)
)
2
where LH(i) is an LH subband at level i, W is a wavelet coefficient, m is a coefficient row, n is a coefficient column, M is equal to a total number of coefficient rows, N is equal to a total number of coefficient columns, and μ is a corresponding coefficient mean absolute value.
3. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said image parameters including a mean absolute value texture angle and a variance value texture angle, said mean absolute value texture angle being calculated according to a formula:
θ
μ
(
i
)
=
arctan
μ
LH
(
i
)
μ
HL
(
i
)
where θ μ(i) is said mean absolute value texture angle, μ is a coefficient mean absolute value, i is a subband level, LH is a Low-High subband, and HL is a High-Low subband.
4. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said image parameters include a mean absolute value texture angle and a variance value texture angle, said variance value texture angle being calculated according to a formula:
θ
σ
(
i
)
=
arctan
σ
LH
(
i
)
σ
HL
(
i
)
where θ σ(i) is said variance value texture angle, σ is a coefficient variance value, i is a subband level, LH is a Low-High subband, and HL is a High-Low subband.
5. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said image parameters including total mean absolute values that are calculated according to a formula:
μ (i) =[μ LH(i) 2 +μ HH(i) 2 +μ HL(i) 2 ] 1/2
where μ (i) is one of said total mean absolute values, i is a wavelet level, μ LH(i) is a first coefficient mean absolute value for an LH subband, μ HH(i) is a second coefficient mean absolute value for an HH subband, and μ HL(i) is a third coefficient mean absolute value for an HL subband.
6. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said image parameters including total variance values that are calculated according to a formula:
σ (i) 2 =σ LH(i) 2 +σ HH(i) 2 +σ HL(i) 2
where σ (i) is one of said total variance values, i is a wavelet level, σ LH(i) is a first coefficient variance value for an LH subband, σ HH(i) is a second coefficient variance value for an HH subband, and σ HL(i) is a third coefficient variance value for an HL subband.
7. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images, said query image being a texture model of a composite image that is formed by merging a series of homogeneous sample images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, one of said distance values for said texture model and one of said test images being calculated according to a formula:
D
=
∑
i
=
1
4
1
2
i
[
μ
(
i
)
T
θ
μ
(
i
)
T
-
θ
μ
(
i
)
I
+
1
5
σ
(
i
)
T
θ
σ
(
i
)
T
-
θ
σ
(
i
)
l
]
where D is said one of said distance values, T indicates said texture model,
I indicates said one of said test images, i is a wavelet level, μ is a total mean absolute value, σ is a total variance value, θ σ is a variance value texture angle, and θ μ is a mean absolute value texture angle.
8. A system for performing an image search procedure with an electronic device, comprising:
image data stored in a memory device that is accessible to said electronic device, said image data including a query image and test images; and
a feature detector that calculates image parameters corresponding to said image data, said feature detector utilizing said image parameters to calculate distance values that represent texture similarity characteristics between said query image and said test images, said feature detector evaluating said distance values to determine a matching image from among said test images, said feature detector performing a texture/color model based retrieval procedure by initially filtering out all undesired color information from said query image to produce a filtered color image, said feature detector then converting said filtered color image into a four-level wavelet transform with color coefficients, said feature detector also filtering out all color information from said query image to produce a luminance image, said feature detector then creating a four-level Haar wavelet transform from said luminance image to produce wavelet coefficients.
9. The system of claim 8 wherein said feature detector performs a pointwise product procedure to produce a composite texture/color model for performing said texture/color model based retrieval procedure, said pointwise product procedure being performed by multiplying each of said color coefficients with corresponding ones of said wavelet coefficients to produce composite texture/color coefficients.
10. The system of claim 9 wherein said texture/color retrieval procedure includes calculating a coefficient mean absolute value μ for a subband LH by utilizing a formula:
μ
L
,
H
(
i
)
=
∑
m
=
1
M
∑
n
=
1
N
C
LH
(
i
)
[
m
,
n
]
W
LH
(
i
)
[
m
,
n
]
∑
m
=
1
M
∑
n
=
1
N
C
LH
(
i
)
[
m
,
n
]
where LH(i) is said subband LH at level i, W is a wavelet coefficient, C is a filtered color coefficient, m is a coefficient row, n is a coefficient column, M is equal to a total number of coefficient rows, and N is equal to a total number of coefficient columns.
11. The system of claim 9 wherein said texture/color retrieval procedure includes calculating a coefficient variance value σ for a subband LH by utilizing a formula:
σ
LH
(
i
)
2
=
∑
m
=
1
M
∑
n
=
1
N
(
C
LH
(
i
)
[
m
,
n
]
W
LH
(
i
)
[
m
,
n
]
-
μ
LH
(
i
)
)
2
∑
m
=
1
M
∑
n
=
1
N
C
LH
(
i
)
[
m
,
n
]
where LH(i) is said subband LH at level i, W is a wavelet coefficient, C is a filtered color coefficient, m is a coefficient row, n is a coefficient column, M is equal to total number of coefficient rows, N is equal to total number of coefficient columns, and μ is a corresponding coefficient mean absolute value.Cited by (0)
No later patents cite this yet.
References (0)
No backward citations on record.